CLI for Coding Agents

Wallaby CLI brings runtime intelligence into agentic coding workflows. Whether you’re using Claude Code, Codex CLI, Copilot CLI, OpenCode, Pi, or any other AI coding agent, Wallaby can now run directly from the command line without requiring an editor session or an MCP server to be running. This is especially useful for git worktrees and other headless workflows.

Adding Wallaby CLI support to a coding agent is straightforward. You can enable it with npx skills add https://github.com/wallabyjs/skills --skill wallaby-cli. Wallaby starts only when an agent actually needs it, stops automatically when the agent exits, and reuses an existing instance started from an editor session when available. This minimizes startup time and provides immediate access to runtime data.

With Wallaby CLI, coding agents can use accurate test and coverage data to validate generated code, improve test quality, and increase coverage. Agents can also debug failing tests without modifying source files, reducing unnecessary edits and making iterative workflows more efficient and token-effective.

This is an early step, and there’s more coming. We’re continuing to improve support for isolated, sandboxed, and CI environments. We’re also working on connecting existing editor extensions to CLI-managed Wallaby instances, improving the skill’s token efficiency through evals and other DX enhancements, and exposing even more Wallaby runtime data to agents through the CLI.